linux-kernel - Re: [RFC PATCH v4 7/8] xfs: reserve blocks for truncating realtime inode

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20240531152922.GN52987@frogsfrogsfrogs>
Date: Fri, 31 May 2024 08:29:22 -0700
From: "Darrick J. Wong" <djwong@...nel.org>
To: Christoph Hellwig <hch@...radead.org>
Cc: Zhang Yi <yi.zhang@...weicloud.com>, linux-xfs@...r.kernel.org,
	linux-fsdevel@...r.kernel.org, linux-kernel@...r.kernel.org,
	brauner@...nel.org, david@...morbit.com, chandanbabu@...nel.org,
	jack@...e.cz, willy@...radead.org, yi.zhang@...wei.com,
	chengzhihao1@...wei.com, yukuai3@...wei.com
Subject: Re: [RFC PATCH v4 7/8] xfs: reserve blocks for truncating realtime
 inode

On Fri, May 31, 2024 at 07:13:05AM -0700, Christoph Hellwig wrote:
> On Fri, May 31, 2024 at 07:10:00AM -0700, Darrick J. Wong wrote:
> > On Fri, May 31, 2024 at 05:42:37AM -0700, Christoph Hellwig wrote:
> > > > -	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_itruncate, 0, 0, 0, &tp);
> > > > +	resblks = XFS_IS_REALTIME_INODE(ip) ? XFS_DIOSTRAT_SPACE_RES(mp, 0) : 0;
> > > 
> > > This probably wants a comment explaining that we need the block
> > > reservation for bmap btree block allocations / splits that can happen
> > > because we can split a written extent into one written and one
> > > unwritten, while for the data fork we'll always just shorten or
> > > remove extents.
> > 
> > "for the data fork"? <confused>
> > 
> > This always runs on the data fork.  Did you mean "for files with alloc
> > unit > 1 fsblock"?
> 
> Sorry, it was meant to say for the data device.  My whole journey
> to check if this could get called for the attr fork twisted my mind.

I really hope not -- all writes to the attr fork have known sizes at
syscall time, and appending doesn't even make sense.

> But you have a good point that even for the rt device we only need
> the reservation for an rtextsize > 1.

<nod>

--D